Friday, 18th of October (2018)

The Genetics of T2D and Associated Traits

Marullo, El-Sayed Moustafa, & Prokopenko (2014)

A lot of SNPs discovered, but …

Flannick & Florez (2016)

Glycaemia and Type 2 Diabetes

A Genetic Joint Effect?

Goals & Rationals

"In all cases, the glucose-raising allele was associated with increased risk of T2D, yet fasting glucose effect sizes and T2D ORs were weakly correlated"

Goals & Rationals

Scott et al. (2012)

Goals & Rationals

  • Goals
    Find SNPs associated with an elevated glycaemia and T2D risk

  • Data (Metabochip)
    D.E.S.I.R.

  • Method
    Joint Model

Joint Model

Overview

Joint Model

Overview

adapté de Ibrahim, Chu, & Chen (2010)

Joint Model

Overview

Ibrahim et al. (2010)

Joint Model

Why?

Clinical studies goals:

* to identify biomarker relavant to a disease;
* TO identify the effect of a treatment on a disease (independently of the association with the biomarker).

Example:

* __Biomarker__ fasting glucose
    * __Even__ Type 2 Diabetes

Joint Model

Cox Model?

Cox model with time-varying covariate: \[\begin{align} h(t)=h_0(t) \exp(\beta Y(t) + \alpha Z) \end{align}\]

Joint Model

Limitations of the Cox Model

  • Biomarker are mostly not measures at all times and/or event time.
    • Biormarkers are measured at determined time points (\(t_{ij}\))
    • Biormarkers can have missing values over time * \(\Rightarrow\) Imputation? \(\Rightarrow\) Bias introduction
      • Biormarkers are measured with some degree of error
      • \(\Rightarrow\) Noise in the biomarker trajectory (\(Y_i(t_{ij}) \neq X_i(t_{ij})\))
    • Biomarker can be endogenous
      • \(\Rightarrow\) Biomarker's trajectory can change when the event occurs \(\Rightarrow\) Bias introduction

Modèle Joint

Avantages

  • Tester l'effet d'un SNP simultanément sur:

    • le biomarqueur (\(\gamma\));
    • la survenue de l'événement (\(\alpha\));
    • le biomarqueur et la survenue de l'événement (\(\beta\gamma+\alpha\)).
  • Gain de puissance statistique:

    • si \(\beta\neq0\) (pour détecter un effet conjoint du SNP \(\beta\gamma+\alpha\neq0\)) ;
    • par rapport à un modèle de "Cox étendu" (modèle de Cox avec covariable dépendante du temps).

Modèle Joint

Formulation

Dans le Modèle Joint, deux composantes (modèles) sont utilisées:

* un modèle longitudinal est utilisé pour modéliser la "vraie" trajectoire du biomarqueur (non-observable),
* et est incorporé en tant que covariable (latente) dépendante du temps dans un modèle de survie.

Modèle Joint

Formulation: Composante Longitudinale

Composante longitudinale: modèle (linéaire) mixte

Avec \(T_i\), le temps d'événement et \(C_i\), le temps de censure à droite: \[\begin{align}\tilde{T_i}&=\min(T_i, C_i),\end{align}\]

avec \(t_{ij}\leq \tilde{T_i}\), \(j=1, \cdots, n_i\), où \(n_i\) est le nombre de mesures.

Pour un individu \(i\): \[\begin{align}Y_{i}(t_{ij})=X_{i}(t_{ij})+\epsilon_{i}(t_{ij})\end{align}\]

Où: \[\begin{align}\epsilon_{i}(t_{ij}) \sim \mathcal{N}(0, \sigma^2)\end{align}\]

Modèle Joint

Formulation: Composante Longitudinale

Composante longitudinale: modèle (linéaire) mixte

Pour un individu \(i\): \[\begin{align}Y_{i}(t_{ij})=X_{i}(t_{ij})+\epsilon_{i}(t_{ij})\end{align}\]

La trajectoire \(X_i(t_{ij})\) peut être définie par un polynome, fonction du temps \(t_{ij}\): \[\begin{align} X_{i}(t_{ij})=\theta_{0i} + \theta_{1i}t_{ij} + \cdots + \theta_{pi}t_{ij}^p &, & \boldsymbol\theta_p &\sim \mathcal{N}(\boldsymbol\mu_, \boldsymbol\Sigma) \end{align}\]

Exemple: D.E.S.I.R. Avec \(\W_i\), la matrice des covariables d'ajustement (^Age, Sexe et IMC) et \(Z_i\), le génotype du SNP d'intérêt: \[\begin{align} Y_{i}(t_{ij})=\theta_{0i} + \theta_{1i}t_{ij} + \gamma Z_i + \delta W_i + \epsilon_{i}(t_{ij}) \end{align}\]

Modèle Joint

Formulation: Composante de Survie

Composante de survie: modèle de Cox à risque proportionnel

Interrelation entre \(X_i(t)\), \(T_i\) et \(W_i\): \[\begin{align} h_i(t)&=\lim_{dt \to 0} \frac{P\{t\leq T_i<t+dt|T_i\geq t, \bar{Y_i}(t), W_i\}}{dt}\\ &=h_0(t) \exp\{\beta X_{i}(t) + \eta W_i\} \end{align}\]

Où \(\bar{Y_i}(t)=\{Y_i(u),0 \leq u \leq t\}\) est l'historique de la trajectoire jusqu'au temps \(t\).

Exemple: D.E.S.I.R. Avec \(W_i\), la matrice des covariables d'ajustement (^Age, Sexe et IMC) et \(Z_i\), le génotype du SNP d'intérêt: \[\begin{align} h_i(t_{ij})&=h_0(t_{ij}) \exp\{\beta X_{i}(t_{ij}) + \alpha Z_i + \eta W_i\} \end{align}\]

Modèle Joint

Test d'Hypothèse

Pour tester l'hypothèse nulle: \[\begin{align}H_0&:& \theta=\theta_0\\ H_1&:& \theta\neq\theta_0\end{align}\]

  • Test du Rapport de Vraisemblance \(LRT=-2\{\ell(\hat{\theta}_0)-\ell(\hat{\theta})\}\)

  • Test de Wald \(W=(\hat{\theta}-\theta_0)^\top \mathcal{I}(\hat{\theta})(\hat{\theta}-\theta_0)\) (Univarié: \((\hat{\theta}_j-\theta_{0j})/\widehat{s.e.}(\hat{\theta}_j)\))

  • Test du Score \(U=S^\top(\hat{\theta}_0)\{\mathcal{I}(\hat{\theta}_0)\}^{-1}S(\hat{\theta}_0)\)

Modèle Joint

Etude de Simulation: Objectifs

  • Précision des estimateurs: RMSE (Root-Mean Square Error)

Root-Mean Square Error \[\begin{align} \operatorname{MSE}(\hat\theta)&= \operatorname{Biais}(\hat\theta)^2 + \operatorname{Var}(\hat\theta)\\ \operatorname{RMSE}(\hat{\theta})&=\sqrt{\operatorname{MSE}(\hat\theta)}\\ &=\sqrt{E\{(\hat{\theta}-\theta)^2\}} \end{align}\]

  • Faisabilité à l'échelle du génome: temps de calcul

Modèle Joint

Etude de Simulation: Objectifs

  • \(\Rightarrow\) Comparaison à une approche en "deux-étapes" (TS) :

"Two-Step" * Première étape:

\[\begin{align} Y_{i}(t)&=X_i(t)+ \epsilon_{i}(t)\\ X^*_{i}(t)&=E\{X_{i}(t)|\bar{Y_i}(t), T_i\geq t\} \end{align}\]

  • Deuxième étape:

\[\begin{align} h_i(t)=h_0(t) \exp\{\beta X^*_{i}(t)\} \end{align}\]

Modèle Joint

Etude de Simulation: Génération des Données

Modèle longitudinale: \(Y_{i}(t)=\theta_{0i} + \theta_{1i}t + \gamma Z_i + \epsilon_{i}(t)\)

Modèle de survie: \(h_i(t)=h_0(t) \exp\{\beta X_{i}(t) + \alpha Z_i\}\)

Génération des temps d'événement: distribution exponentielle \[\begin{gather} \lambda>0\\ h_0(t)=\lambda\\ H_0(t)=\lambda t\\ H_i(T_i)=\int_0^{T_i}\lambda \exp(\beta X_i(t)+\alpha Z_i)dt\\ T_i=\frac{1}{\beta\theta_{1i}}\log\left(1-\frac{\beta\theta_{1i}\times \log(1-u)}{\lambda \exp(\beta\theta_{0i}+(\beta\gamma+\alpha)Z_i)}\right),\ u\sim\mathcal{U}(0,1)\\ \end{gather}\]

Modèle Joint

Etude de Simulation: Paramètres de Simulation

Taille d'effet \(\beta\) et Odd Ratio respectivement pour la concentration de glucose et le diabète de type 2.

Modèle Joint

Etude de Simulation: Paramètres de Simulation

Me?

Flannick, J., & Florez, J. C. (2016). Type 2 diabetes: Genetic data sharing to advance complex disease research. Nature Reviews Genetics, advance online publication. https://doi.org/10.1038/nrg.2016.56

Ibrahim, J. G., Chu, H., & Chen, L. M. (2010). Basic Concepts and Methods for Joint Models of Longitudinal and Survival Data. Journal of Clinical Oncology, 28(16), 2796. https://doi.org/10.1200/JCO.2009.25.0654

Marullo, L., El-Sayed Moustafa, J. S., & Prokopenko, I. (2014). Insights into the Genetic Susceptibility to Type 2 Diabetes from Genome-Wide Association Studies of Glycaemic Traits. Current Diabetes Reports, 14(11). https://doi.org/10.1007/s11892-014-0551-8

Scott, R. A., Lagou, V., Welch, R. P., Wheeler, E., Montasser, M. E., Luan, J., … Barroso, I. (2012). Large-scale association analyses identify new loci influencing glycemic traits and provide insight into the underlying biological pathways. Nature Genetics, 44(9), 991–1005. https://doi.org/10.1038/ng.2385